5 Comments

Summary:

Amazon Web Services is adding a flexible IOPS storage option to its Relational Database Service. People setting up new MySQL, Oracle or SQL Server instances can take advantage of the new option now. Later, they can move legacy instances over.

Amazon Web Services
photo: Flickr/Will Merydith

Not all database workloads are created equal. Some high-priority jobs require faster storage input/output operations per second than others. Now Amazon is acknowledging that fact with new provisioned IOPs for its Relational Database Service.

IOPs are basically the round trips data takes between database and storage. Amazon started offering analogous provisioned IOPS for its Elastic Block Storage (EBS), but is now expanding that capability to all of its supported databases, according to the Amazon Web Services blog. According to the blog:

Effective immediately, you can provision new RDS database instances with 1,000 to 10,000 IOPS, and with 100GB to 1 TB of storage for MySQL and Oracle databases. If you are using SQL Server, the maximum IOPS you can provision is 7,000 IOPS. All other RDS features including Multi-AZ, Read Replicas, and the Virtual Private Cloud, are also supported.

Note: developers will have to provision their maximum IOPS when they start out — and can scale up to that maximum over time.  If they want to exceed their maximum they will have to start again.

Later on, users will be able to move existing legacy RDS instances to provisioned IOPS storage — at least for MySQL and Oracledatabases. If those who just can’t wait, they can take existing data, export it and re-import it into a new database instance, according Amazon.

Amazon is known for releasing  streams of new and enhanced features over time. That pace will likely continue or even hasten as more cloud competitors — Rackspace, HP, et al — come online. Disk IO can be an issue for Amazon EC2 customers. One account told me this week that they’d moved some of their high I/O tasks from Amazon to Rackspace for this very reason.

For more, check out the video.

  1. Disk performance on EC2 has always been an issue. A year ago I remember talking to people using EBS who found that performance would randomly drop off and could only be restored by destroying the volume and recreating it. The problem was not so much the performance levels themselves, but the variability. This is a problem for i/o bound applications i.e. databases.

    Amazon released this first as IOPS optimised instance types but seems they’re now rolling it out to their more “managed” database-as-a-service products. The guys at Parse seem to have had a good experience with this so far – http://blog.parse.com/2012/09/17/parse-databases-upgraded-to-amazon-provisioned-iops/

    Share
    1. thanks @david. i meant to include a sentence to that effect.

      Share
  2. Excited to see this feature! We have run a cost analysis and it turns out quite expensive. For our deployment, the provisioned IOPS takes up around 46% of the total cost. Shared our experiment results here: http://blog.planforcloud.com/2012/09/rds-iops-expensive-high-performance.html

    Share
    1. That’s really helpful.

      Share
  3. Folks at Amazon Web Services are learning fast and offering a range of granular services to fill up all business needs. They are improving on their initial understanding of the market. What is the probability that they’d will be a disruptive force against the traditional storage industry?

    Share

Comments have been disabled for this post